Nvidia (Mellanox) SN3700 all ports dead: Diagnose & Fix
By Sai Kiran Pandrala · reviewed by Sai Kiran Pandrala, Editor Last verified: 2026-05-30
| Vendor | Nvidia (Mellanox) |
|---|---|
| Operating system | Cumulus Linux / NVOS / SONiC |
| Category | Hardware Failure |
| Skill level | Intermediate to advanced |
| DIY-able? | Yes with CLI access; some scenarios need Nvidia Enterprise Support + RMA. |
Across years of operating Nvidia (Mellanox) gear I have watched the same hardware-failure pattern repeat: a unit ships fine, runs for two years, then trips on a power-event or a thermal excursion. On Cumulus Linux / NVOS / SONiC the recovery path is the same whether the affected unit is from the SN3700 family or something newer.
Before you touch anything, capture state. `nv show system` and `nv show platform environment` dumped to a file is worth more than a screen-cap because Nvidia Enterprise Support will ask for the exact output when you open the case. Keep the artifact even if the box recovers on its own.
Below I walk through the on-box steps first, then the Nvidia Enterprise Support escalation path. If you have spares on hand, swap-then-diagnose is usually faster than diagnose-then-swap, but only if you can afford the rack time.
What this guide covers
Diagnose and recover from all ports dead on a Nvidia (Mellanox) SN3700.
Step-by-step
- Try the same cable + endpoint on a known-good port to confirm the issue is the device.
- If modular, re-seat the affected line card.
- Check the platform / hardware status command.
- If a single line card is dead, RMA it. If the supervisor or chassis, RMA accordingly.
CLI / commands
# Verify hardware state
nv show system
nv show platform inventory
nv show platform environment
# Collect for Nvidia Enterprise Support
cl-support (Cumulus) / show techsupport (SONiC)
When to RMA
- Repeated failure after re-seat and power-cycle
- Visible burn, scorching, or physical damage
- POST or memory diagnostic failure
- Hardware crashinfo without a software workaround
Frequently asked questions
Will this work on my specific Cumulus Linux / NVOS / SONiC version?
The procedure reflects current Cumulus Linux / NVOS / SONiC behaviour. Older releases may need minor syntax adjustments: use the CLI help (? or tab-completion) to verify.
Should I open a Nvidia Enterprise Support case immediately?
Open one if you suspect hardware failure or the symptom persists after a maintenance-window reload. Make sure your support entitlement is active first.
Where can I find the Nvidia (Mellanox) official documentation?
https://docs.nvidia.com/networking/, search the product family + feature name.
Is this procedure safe in production?
Test in a lab or maintenance window first. Capture pre-change state so you can roll back.
Related guides
Related fixes
Related guides worth a look while you sort this one out:
- Nvidia (Mellanox) SN2010 all ports dead: Diagnose & Fix
- Nvidia (Mellanox) SN2100 all ports dead: Diagnose & Fix
- Nvidia (Mellanox) SN2410 all ports dead: Diagnose & Fix
- Nvidia (Mellanox) SN2700 all ports dead: Diagnose & Fix
- Nvidia (Mellanox) SN3420 all ports dead: Diagnose & Fix
- Nvidia (Mellanox) SN3700 single port dead: Diagnose & Fix
References
- Nvidia (Mellanox) support portal: https://enterprise-support.nvidia.com/
- Nvidia (Mellanox) knowledge base: https://docs.nvidia.com/networking/
- Nvidia (Mellanox) security advisories: https://www.nvidia.com/en-us/security/
- Open a case: https://enterprise-support.nvidia.com/s/createcase
Reference material, not professional advice. Validate against your specific Cumulus Linux / NVOS / SONiC version and test in a non-production environment before applying.
What changed recently?
Fault diagnosis on a Nvidia device goes faster when you map the symptom to a recent change:
- Did firmware update in the last 7 days?
- Did the network (router, ISP, VPN) change?
- Was the device moved physically?
- Did paired devices (phone, hub, app) update?
- Were any accessories swapped in or out?
The answer narrows the root cause to a manageable subset.
Before you start
A few things to confirm so the Nvidia device fix goes cleanly:
- Latest firmware downloaded if you're going to update.
- Warranty + support contract status checked. opening sealed parts may void it.
- Backup of current configuration (where applicable) taken.
- Spare parts on hand if you anticipate replacement.
- Adequate workspace, lighting, and time, rushing causes regressions.
How to confirm it's actually fixed
On a Nvidia device, the test is rarely "reboot and see". Use this list:
- Active reproduction: trigger the original failure path on purpose.
- Indirect reproduction: do an activity that would expose the same subsystem.
- Status indicator review: every LED / display / app status should be green.
- 24-hour soak: leave the device under normal load overnight; check the next morning.
- Telemetry check: review the device or app's diagnostic log for new error entries.
Escalation guide
For a Nvidia device, the right escalation depends on impact:
- Cosmetic / minor: log a ticket via the Nvidia app or web portal. Response 1-3 business days.
- Mid-impact: phone support. Have your serial number ready.
- Critical (production down, safety issue): in-person dealer / TAC visit. Bring proof of purchase.
- Out of warranty: third-party repair shop with manufacturer-certified technicians.
More frequently asked questions
Why is this happening on a brand-new unit?
Out-of-box defects do occur. If you've owned the device under 30 days and the symptom persists after a factory reset, escalate to the seller for replacement under DOA terms before opening a manufacturer support case.
What if my model isn't exactly the same revision?
Cross-check the model code on the rating plate against the manufacturer support page. Major firmware generations sometimes shift the menu path; the option is usually under a similarly-named section.
Is it safe to apply during business hours?
If the device is in production use, apply during a scheduled maintenance window. Most procedures need 2-15 minutes of downtime. Capture pre-change state so you can roll back if needed.
How long does this fix usually take?
Most users complete the steps in 20-45 minutes the first time, and 5-10 minutes on subsequent runs once the menu paths are familiar.
Should I update firmware first or last?
Update firmware first if a release note specifically mentions your symptom. Otherwise, finish the troubleshooting flow first, then update; that way you can isolate whether the update or the underlying fix solved it.